Introduction

The national political media was captivated in the summer of 2022 when the reality TV star and famous New Jerseyan known as Snooki recorded a short video wishing luck to Pennsylvania Senate candidate Mehmet Oz (R) on his recent (but, in her words, “temporary”) move to Pennsylvania from New Jersey. Oz’s opponent, then-Lt. Governor John Fetterman (D) had paid for the video to highlight Oz’s permanent residence in New Jersey, despite running to represent neighboring Pennsylvania in the Senate. With this and other unique tactics, the Fetterman campaign continually broadcasted his own authentic, place-based connection to his home state (Fetterman famously tattooed his hometown’s zip code on his arm), while lambasting his opponent as an out-of-touch “carpetbagger.” Despite the poor national political environment for Democrats in 2022, as well as the serious stroke Fetterman suffered in May, the Pennsylvania-centric strategy ended up working. Fetterman outperformed the polling, the bad press, and the national political fundamentals, winning by nearly 5% of the vote.

The sense that local connections matter has to be considered alongside compelling recent scholarship demonstrating a persistent nationalization of American politics. Clustered and polarized “mega identities” centered around the two parties and their supportive social groups have become primary determinants of many Americans’ political attitudes and vote choices (Green et al. 2002, Mason, 2018, Sides et al., 2019). The result has been growing loyalty to nationalized party brands and voting choices that are increasingly determined by these allegiances (Abramowitz & Webster, 2016). Likewise, voters and the media are more focused on issues generated by national figures, and issues that are of national significance (Hopkins, 2018).

Other important recent work, however, has shown that local, place-based identities and connections of the kinds that Fetterman shares with his fellow Pennsylvanians remain uniquely suited to cut across partisan boundaries in shaping voter behavior. Recent experimental studies have demonstrated these place-based connections between voters and candidates (Campbell et al. 2019; Jacobs & Munis 2018; Munis, 2021; Panagopoulos et al., 2017; Schulte-Cloos & Bauer, 2021), and observational studies have found local-candidate electoral advantages some limited contexts (Hunt 2022a; Hunt, 2021a, 2021b; Meredith, 2013).

In this paper, we build on this work by considering the breadth and depth of these effects across multiple domains and using different methodological tools. We analyze the impact of local candidate roots in general elections for the U.S. Senate, an office that has been largely ignored by the literature. Our primary analysis comes from a nationwide survey featuring a randomized controlled experiment (N =  ~ 2100), in which we find that Senate candidates with place-based backgrounds in their home states enjoy general election advantages ranging on average from 2 to 7 percentage points of vote share, even after accounting for partisanship, incumbency, and other major factors in elections to the U.S. Senate. We supplement these findings with comprehensive observational data from more than 60 years of major-party candidates for the U.S. Senate (1960–2020, N =  ~ 1200), the analysis of which largely mirrors the experimental findings.

We also uncover new evidence on the groups of voters most responsive to local candidate ties. Specifically, we find that the electoral advantage associated with them draws largely from voters supporting locally-rooted candidates of the opposing party, suggesting that the electoral benefits associated with local place identity are cross-cutting in nature. Relatedly, both our experimental and observational findings suggest that local candidate effects are asymmetrical between the parties, with Democratic candidates benefitting more, and in different ways, from local connections. Finally, we cite descriptive observational evidence that local candidate advantages have not substantially diminished in the U.S. Senate over the past sixty years, indicating that the reemergence of political nationalization and partisan polarization in recent decades has done little to temper voters’ appetites for homegrown candidates.

Local Ties, Electoral Advantages, and Representation

Understanding whether and how local connections matter to voters is crucial given the centrality of place to how political representation is intended to operate in the United States. The framers of the American Constitution created a federal system in which representatives were selected based on their geographic location. The Federalist Papers argued in favor of an “intimate sympathy” between the voters in a place and candidates connected to that place (Madison, 1789). Because places would define voters, places would be the means used to organize voters. The Constitution reflects this sentiment by requiring that members of Congress be an “inhabitant” of the state they are running in at the time they take office. Regardless of whether one agrees with these features normatively, it was and remains the way our federal system of elections and representation operates under the law. And although the United States is not unique in this regard, it may be that the vastness of the country, combined with its geographic diversity calls for more particularistic place representation from elected leaders.

These fundamentals of American political representation are the backdrop for a consistent finding in political science research: voters’ tendency to reward candidates who have biographical connections to the geography of their particular jurisdictions; and to punish those who lack these connections. Early work explored what V.O. Key (1949) called the “friends-and-neighbors” electoral effect, in which candidates for statewide office outperform expectations in counties where they live, and in adjacent areas. Likewise, Fenno (1978) provided a breadth of qualitative evidence that members of Congress in particular place significant value on their local ties, seeing them as crucial to their electoral success, and deeply impactful on their behavior in their districts.

Since these initial findings, however, American politics has nationalized on a number of fronts. The two parties have sorted more cleanly into two distinct cultural and ideological factions in the electorate (Levendusky, 2009; Mason, 2018) and in institutions like Congress (Hopkins, 2017; Theriault, 2008; Thomsen, 2014). The parties have become more homogeneous across space, and reflect nationalizing trends in campaigns and elections, including campaign financing (Canes-Wrone & Miller, 2021). Political media has been nationalizing as well, with a significant decline in local news and an outsized focus on national political developments (Hayes & Lawless, 2021; Moskowitz, 2021). It has also been undergoing a parallel fragmentation, giving Americans greater ability to fit their news consumption into nationalized partisan frames (Mancini, 2013). These nationalization trends extend to voter behavior. Candidates for federal offices like the House and Senate, state offices like governor and state legislature, and even many local offices have come to be judged by the voters the same nationalized partisan criteria (Abramowitz & Webster, 2016; Amlani & Algara, 2021; Zingher & Richman 2019), perhaps leaving less room for local considerations.

Amidst these nationalizing trends, an emerging set of modern scholarship has replicated Key’s and Fenno’s original findings, demonstrating local candidate preferences in a variety of electoral contexts and jurisdictions (Campbell et al., 2019; Gimpel et al., 2008; Hunt 2022b; Meredith, 2013; Panagopoulos et al., 2017; Schulte-Cloos & Bauer, 2021; Tatalovich, 1975). Crucially, these represent within-jurisdiction effects—for example, U.S. Senate candidates out- or under-performing in some parts of their state versus other parts. Recent works by Hunt (2021a, 2021b, 2022a, 2022b) and Munis (2021), however, have established that local roots can have meaningful jurisdiction-wide effects on both the primary and general election fortunes of actual or hypothetical candidates for office (i.e., having local roots in one’s district as a whole, versus not having them). Many of the studies demonstrating these effects have done so in electoral systems outside the United States (Evans et al., 2017; Tavits, 2010), including in more party-centric (that is, less candidate-centric) electoral systems (Fiva & Halse, 2016; Fiva & Smith, 2017; Jankowski, 2016).

These studies have also theorized the mechanisms by which local candidate attributes encourage voters to support them, and perhaps also to punish their “carpetbagging” counterparts. There are generally two different but complementary mechanisms that have been used to explain these effects.

The first highlights practical reasons for local candidate preferences. For instance, deeply-rooted candidates benefit from higher local name recognition as a result of spending more time in the area over the course of their lives. Locals also have more time to develop robust social, political, and economic networks that can be useful in campaigns. They have longer-term access to authentic, hard-won knowledge of the issues, ideologies, and political cultures that define their jurisdiction and the constituents who inhabit it. This puts so-called “carpetbaggers” at an information disadvantage, and more prone to outsider-status gaffes on the campaign trail. This was a common issue for the previously discussed example from Pennsylvania: Mehmet Oz was mercilessly mocked for mixing up the names of local Pennsylvania grocery store names in a widely-viewed campaign video. Importantly, these more instrumental mechanisms for the local roots advantage can happen whether voters are specifically aware of a candidate’s biographical ties to the area or not.

A second mechanism is the symbolic representational connection that shared local roots can forge between candidates and voters. Earlier locally-focused areas of representation research emblemized by Fenno (1978) and others (e.g. Parker & Goodman, 2009) considered the shared nature of local roots and biographical backgrounds as a way of demonstrating that a candidate is “one of” the voters they are running to represent, and will therefore have their best interests at heart because they are like them, much as a Black representative can identify with their co-racial constituents. Fenno in particular emphasized place and local authenticity as central to the reputations of many members of Congress. In these cases, the deep place-based roots that many of Fenno’s subjects possessed gave them opportunities to showcase these roots authentically via their home style behaviors in the district—another advantage locally-rooted candidates have over their unrooted counterparts.Footnote 1

More recent work has elaborated on these symbolic connections with the concept of “place identity”, in which voters imbue their home places with personal sentiment, self-worth, and political meaning (Childs and Crowley 2011; Cramer, 2016; Munis, 2022; Sajuria & Collignon, 2018,). The result is a clear electoral preference for local candidates who share this place identity (Horiuchi et al., 2018; Munis, 2021). This is consistent with related findings on “context” effects showing that voters draw information from their local environments to inform their political behaviors (Enos, 2017; Newman et al., 2015).

One important element of these symbolic benefits has been largely overlooked by previous work: the unique capacity of local ties to garner cross-party appeal from voters. We posit that this appeal is a major contributor to local roots' electoral power. Local connections, rather than simply taking their place alongside other polarizing identities, are well-suited to cross-cut nationalized party attachments and win over opposite-party voters. Nationalized identities like party, ideology, and race are powerful in part because they can be shared among members of the in-group regardless of where they live. The same is true of ideological attachments: as Mason (2018) and others have pointed out, conservatism and liberalism as ideological identities increasingly mean the same thing to those voters who possess that identity nearly regardless of where they live.

Meanwhile, local identities based on shared characteristics like biographical roots can also generate attachments and political support, but in a way that is universal within the jurisdiction. A Senate candidate’s deep local roots in Missouri offers them a shared identity not with any nationalized political faction, but—by definition—with Missourians in particular. Indeed, being “a Missourian” is the only descriptive trait that this candidate and all of their future constituents have in common, regardless of whether they share the same party or ideological orientation. Candidates with deep local ties in Missouri, therefore, can leverage and emphasize this descriptive jurisdiction-wide bond with great effectiveness. As a result, possession of local roots does not require sacrificing other key identities. For example, a Republican hoping to appeal to independent or Democratic voters may moderate their ideological stances, but in doing so risks alienating their staunchly conservative supporters in a primary. Local roots, on the other hand, are theoretically appealing to any voter regardless of party, and come with few electoral risks. Thus, local roots as a candidate attribute that can—and we argue, does—cross-cut nationalized identities like party and ideology.

For all of these reasons, we expect that voters will be inclined to reward Senate candidates with local biographical connections in the form of local roots—that is, being born and/or raised in the state they are running to represent. These rewards, we expect, will come in the form of higher favorability ratings, perceptions of trustworthiness and relatability, and of course their votes.

New Forms and Venues of Evidence

We aim to add to the growing literature demonstrating the power of local candidate connections in a number of substantial ways. First, our contribution spans broader contexts. No study (to our knowledge) has previously analyzed the impacts of statewide roots in the United States Senate.Footnote 2 The Senate provides a useful context for several reasons. Local candidate effects may be unlikely to emerge in such a heavily nationalized political office—possibly even the most nationalized office in American politics outside of the presidency. Voters increasingly select their senators based on their preferences on national issues and in alignment with national candidates (Abramowitz & Webster, 2016, Amlani & Algara, 2021). For instance, very few senators from one party are elected from states that usually prefer candidates from the other party for President.Footnote 3 Even so, the Senate’s uniqueness as an institution also makes it a worthwhile choice for examining place-based connections and representation. For example, state lines and identities tend to be more salient ones for voters, particularly compared to the sometimes arbitrary and incoherent basis of congressional districts. Similarly, its unique status as a multi-member jurisdiction provides further opportunity for same-place candidates to have more varied backgrounds within their states, and relationships with their constituents (Lee & Oppenheimer, 1999; Schiller, 2000).

Second, we test local candidate effects with multiple methodologies that enable us to add important nuance and context to previous findings on local roots. Our primary contribution is a survey experiment fielded in September of 2021 that uses a fictional intervention in actual Senate seats up for grabs in the 2022 midterms to present voters from that jurisdiction with a choice between Democratic and Republican candidates in an open-seat Senate race. We supplement this experiment with an original observational dataset comprising the full cross-section of U.S. Senate races over a six-decade period (1960–2020). This additional context suggests that these findings persist in the real world, and across the broadest possible number of Senate races. The chronological range of the data also permits a descriptive examination of these effects across other time periods.

Local Roots in the Experimental Setting

Our primary analysis tests the effects of local roots using a national panel survey. The respondents for this survey (N = 2100) were drawn from 19 states in which an election for the United States Senate was to be held in 2022.Footnote 4 The respondents reflected a nationally representative sample of voting-eligible adults, with reflective cross-sections of Americans across traits like party, age, gender, race and ethnicity, and geographic area.Footnote 5 In September of 2021, respondents were presented with a fictional news story from a well-known newspaper in their home state reporting that their current incumbent Senator up for reelection in 2022 was not seeking another term.Footnote 6 Respondents were then told that two lesser-known but well-qualified candidates (one from each party) were early frontrunners for their party’s nomination. The treatment then randomly assigned local and national reputations to each of the candidates. A sample of this text can be found below, with the local/national frames highlighted.Footnote 7

[Democrat] Gerald Conley was born and raised in [Florida] and has received much acclaim for his work in the state in various high-profile public and private sector roles. When asked to comment on the open Senate seat, Conley emphasized that “we need a senator who knows us because he is from [Florida] and lives in [Florida]. Only someone who sends his kids to our schools, drinks our water and breathes our air can represent us in Washington.” Conley established a reputation within [Florida] for his vocal opposition to Trump administration policies during President Trump's first term. Conley was also active during the 2020 election, advocating for Joe Biden and other Democrats on the ballot in [Florida].

[Republican] Terry Wilson has never lived in [Florida] previously but has a substantial reputation for his accomplishments in the public and private sector nationally. “[Florida] needs someone representing them who knows how to get things done in this country," Wilson commented following [Sen. Rubio's] announcement of his retirement. “We need someone who can bring that experience to get things done in Washington.” Wilson came to national prominence for his strong advocacy for Donald Trump and other Republicans in the 2020 elections. He also has been a vocal opponent of Biden administration policies throughout 2021.

This framing is appropriate for several reasons. First, the goal was to encapsulate this message in a single fictional article in order to replicate the cumulative effect of an entire campaign message, culminating in candidate evaluations and a hypothetical decision about vote choice. Thus, the goal was to mirror the actual process of the campaign as efficiently as possible. Second, the immediacy of this fictional announcement in the experimental text made it believable to the overwhelming supermajority of our respondents because they might have missed such a recent event.Footnote 8 These fictional news stories were modeled on language from a combination of real-life news articles about local and non-local candidates, and candidate quotations were modeled closely on actual quotations from candidates advertising their local credentials, biographies, and arguments. The design also features typically nationalized partisan primes (such as advocacy for presidential candidates like Joe Biden and Donald Trump) that dominate media coverage and that scholars have shown to impact voter attitudes. The experiment can therefore believably measure how respondents deal with local information in the context in which they will normally receive it—alongside powerful, nationalized partisan frames.

Respondents were debriefed on the fictional nature of the news story immediately following the experiment, and were required to view the debriefing in order to finish the survey. This—coupled with the fact that the survey took place a full 14 months prior to the election in question, and only reached at most a few hundred residents in any given state—gives us confidence that the deception had little to no practical consequences in any of these eventual U.S. Senate elections.Footnote 9

One third of respondents were presented with a local Democrat and a national Republican. A second third were presented with a local Republican and a national Democrat. The third group of respondents was a control group, with all information about local candidate connections removed from the descriptions, but the candidates’ attachments to national partisan figures still in place. Because respondents in each state were randomly assigned to one of these three conditions, the three resulting groups of respondents were comparable in terms of key demographic characteristics like race, gender, and geographic type, as well as their party affiliation and ideological leanings.

Treated respondents consistently favored local candidates compared to the control condition, even in the midst of overt partisan signals in all conditions. We first captured their assessments based on five-point Likert scales that measured how much respondents held favorable views of each candidate; felt they could relate to the candidate; and trusted the candidate. Comparing these candidate-level assessments to each other based on the treatment conditions yielded the expected findings: candidates in the local roots treatment condition consistently outperformed relative to the other conditions on all three measures, particularly on candidate favorability.Footnote 10 More national “carpetbagging” candidates consistently fared worse on these measures.

Fig. 1
figure 1

Average predicted change in probability of voting for the Democratic candidate resulting from the local roots experimental treatments. Error bars represent 95% confidence intervals based on difference of means estimates comparing each treatment condition with the control condition

Full size image
Fig. 2
figure 2

Change in Likert scale evaluations (range from 0–4) of Democratic candidates based on local roots treatments. Error bars represent 95% confidence intervals based on difference of means estimates comparing each treatment condition with the control condition

Full size image
Fig. 3
figure 3

Change in Likert scale evaluations (range from 0–4)of Republican candidates based on local roots treatments. Error bars represent 95% confidence intervals based on difference of means estimates comparing each treatment condition with the control condition

Full size image

Did these assessments carry through to actual predicted vote choices? Here, we averaged the percentage of respondents who said they would support each candidate in the November 2022 general election should they be their party’s nominees. As we see in Fig. 1, the Local Democrat treatment increased the Democratic candidate’s share of the vote by an average of 7 points compared to the control condition.Footnote 11 This effect is statistically significant at the 0.01 level based on difference of means tests comparing the Democratic vote share of respondents in the treatment condition with that of the control condition.

Particularly in electorally competitive U.S. Senate elections, 7 points of vote share is an added advantage that most open-seat Senate candidates would be thrilled to enjoy, and which would almost certainly be decisive in a number of elections (for example, the Fetterman-Oz race described earlier was decided by only a five-point margin). However, Fig. 1 also tells us that the effects of the local candidate treatments are asymmetrical between the parties. Although the presence of a “Local Republican” candidate caused a decrease in Democratic electoral support, the decrease was substantively much smaller and not statistically significant.

What could account for this asymmetry? Although we cannot provide direct evidence for a mechanism, we propose first that—given the significant geographic sorting occurring between the two parties—voters may have come to expect a higher baseline of “rootedness” from Republican (that is, in most instances, more rural) candidates more so than Democratic (urban, more geographically mobile) candidates. This party difference in place conception is reflected in the literature on voter behavior. Jackman and Vavreck (2011), for example, demonstrate the rising role of cosmopolitanism (that is, seeing oneself as a citizen of the world rather than as a local) as a trait predictive of liberal and Democratic votes. In addition, Jacobs and Munis (2018) find evidence that place-based appeals are more effective in rural settings. Thus, Republican respondents (who are generally more rural and less likely to identify as cosmopolitans) were more likely to offer these rewards, particularly when they found them in Democratic candidates, in whom they were “pleasantly surprised” to find an element of localness compared with their possibly stereotypical assumptions about Democrats.

These assumptions from voters and about the parties are reflected in our candidate reputation assessments, as seen in Figs. 2 and 3. They show that the Democratic candidate was rewarded for possessing a trait of local roots that voters have come to expect more from Republicans; meanwhile, he is not punished in the treatment in which he is a carpetbagger, possibly because voters have come to stereotype Democrats in this way due to their generally more cosmopolitan backgrounds. On the other side, the Republican candidate receives modest rewards for a rootedness that voters may already expect him to possess as a baseline trait, while being more significantly punished when he does not possess it.

Heterogeneity and Cross-Cutting Effects

Our experimental data also makes it possible to determine whether the local roots treatment matters differently to different types of voters. Using logistic regressionFootnote 12 to interact the treatment conditions with key subgroup indicators on the dependent variables used above (vote choice, favorability, relatability, and trust), we found that local candidate roots were consistently important across virtually all key subgroups of voters based on individual traits (partisan strength, vote likelihood, race/ethnicity, and educational attainment) as well as contextual factors (geographic neighborhood and state-level partisan lean). We found only three subgroups in which statistically significant differences emerge: unaffiliated votersFootnote 13 were more responsive to the treatments than strong partisans; highly-educated respondents were less responsive; and in some cases, respondents living in swing statesFootnote 14 were slightly less responsive. For the most part, however, candidate localness or non-localness was impactful for virtually every meaningful subgroup of respondents.

We also find novel evidence—in both the candidate assessment and vote choice outcomes—for local candidate roots as a shared connection with voters that cross-cuts party identity in crucial ways. Table 1 displays average treatment effects (in percent change from the control) of local roots on candidate-level assessments like favorability, relatability, and trust, as well as their effects on vote choice (percentage-point increases in vote share relative to the control), split by voters’ party identification. In all treatment conditions, and among all three voter subgroups, local candidate attributes have generally positive effects on all of these outcomes.

Table 1 Local treatment effects on candidate assessments by party identification
Full size table

However, while the literature on nationalized politics suggests that it is increasingly rare to like (Iyengar & Westwood, 2015) or vote for (Abramowitz & Webster, 2016) a candidate of the opposing party, Table 1 also shows that place attachments clearly generate trust, relatability, and even prospective votes from the opposing party that would not otherwise exist. For voters who do affiliate with one of the two major parties, the treatment effects are significantly greater when the opposing party’s candidate is assigned to have deep local roots in their state. Put differently, Democratic voters offer greater local-candidate rewards for Republican candidates than they do for local candidates from their own party, and vice versa.

It is possible that ceiling effects are at work here with same-party support; 96% of Democrat-identifying respondents in the control group reported that they would vote for the Democratic candidate, and 95% of Republicans did the same for their party’s candidate. As a result, there simply may not be much more room for additional same-party support in these samples. However, the even larger cross-partisan effects on the Favorability, Relatability, and Trust measures—which were much more consistent for both Democrats and Republicans—show no evidence of this. Regardless, it is clear that substantial numbers of partisan voters are offering major electoral advantages to locally-rooted opposite-party candidates, and are doing so at even higher rates than they do for candidates in their own party.

Local Roots in the Observational Setting

Respondents’ preference for local candidates—even after accounting for partisanship, ideology, race, and other conditioners of political attitudes as a result of random assignment—was notable in the experimental setting. However, these results may not extrapolate to other contexts, perhaps because it is capturing what voters think they should say rather than what they are actually doing. Do voters respond in the same way in the aggregate when faced with actual candidates, campaigns, and partisan incentives in the real world?

As a supplemental investigation of the findings in the novel survey experiment, we collected observational data on all statewide U.S. Senate elections from 1960 to 2020. This six-decade period is helpful for a few reasons. First, pulling from such a broad period achieves the fullest possible cross-section of candidates, and a variety of political conditions, majority party orientations, and states (particularly since Senate elections are staggered) so as to give us a comprehensive view of what local roots actually look like in the U.S. Senate. Second, it facilitates a descriptive examination of whether the impact of local candidate roots on elections has shifted over time.

Local roots information was gathered using a combination of data from the official Congressional Biographical Directory of the United States and manually-conducted candidate-by-candidate biographical research using historical news archives, candidate biographies, and social media accounts, resulting in successful codes for candidate birthplace for over 95% of open-seat candidates during this time period.Footnote 15 For many of these candidates, particularly those whose candidacy preceded the digital age, additional roots information like high school attendance is largely unattainable. Candidates who actually ended up serving in the Senate, however, had consistent information available about high school, college, and postgraduate school attendance.

Incumbents can therefore offer us a broader picture of what local roots really look like in (successful) candidates for the Senate. Joe Manchin (D-WV), for example, exemplifies the locally-rooted Senator: he was born and raised in Farmington, West Virginia, attended college at West Virginia University, and took over his family’s West Virginia furniture business after college. He also held multiple political offices in West Virginia before his Senate career. Nobody could credibly accuse Manchin of carpetbagging. Contrast his story with that of former Senator (and Secretary of State) Hillary Clinton, who lived in several states before settling in New York just over a year before her first U.S. Senate race.

As Fig. 4 demonstrates, fully one-third of incumbent senators from 1960 to 2020 were born in a state other than the one they represent. More than 40% attended college out-of-state, and more than two-thirds of those with postgraduate degrees attained them in a different state (primarily from top-tier law degree-granting universities). In fact, local high school attendance is the most common local roots attribute for these Senate incumbents. In open-seat races, the numbers are even more surprising: only 44% of all candidates were born in the state they were vying to represent.

Fig. 4
figure 4

Percentage of Senate incumbents who were born, went to high school, undergraduate college, or graduate school in the states they represent in the Senate, 1960–2020

Full size image

Senate candidates with local roots also appear to enjoy the voter-level advantages we observed in the survey experiment, as the descriptive findings in Table 2 indicate. For incumbents, we compiled a Local Roots Index (hereafter LRI) modeled from Hunt (2022a, 2022b) that adds together all local roots components candidates possess. The index ranges from 0 (no early life experience in one’s home state) to 4 (was born, and went to high school, college, and postgraduate school in their home state), and is distributed symmetrically across incumbents during this period.

Table 2 Incumbent electoral conditions based on local roots index score, 1960–2020
Full size table

First, although Senate incumbents are much less likely to run unopposed by the other party than their House counterparts, locally-rooted senators are significantly more likely to do so. They are also less likely to face quality challengers if they are challenged; receive higher shares of the general election vote; and generally more likely to possess prior political experience. Similar descriptive patterns are observed between open-seat Senate candidates who were born in their home state versus those who were not. Of the half of open-seat contenders from 1960 to 2020 who won their races, more than 60% were born in their home state; for those candidates that were defeated, only 27% had that distinction.

The descriptive electoral advantages detailed in Table 2 are mirrored by multivariate analyses on both incumbent and open-seat Senate candidates. Aside from the candidates’ local roots (a “born in-state” indicator for open seats, and the full LRI for incumbents), we also introduce a number of controls traditionally included in congressional election models. First, we include the two-party vote share of the candidate’s party’s presidential nominee in that state in the most recent presidential election to account for statewide partisanship, as well as the candidate’s party’s national popular vote in that year’s U.S. House elections to account for national partisan political conditions.Footnote 16 Second, we include candidate-level variables for race and gender. For open-seat races, we also include candidates’ prior political experience; and for incumbents, we also include institutional attributes like whether the incumbent is a committee chair, chamber leader, or was embroiled in a personal scandal at the time of the election, as well as the current length of their tenure in the Senate. Finally, we include whether the incumbent faced a quality challenger, defined as possession of prior political experience (Jacobson & Carson, 2019).

First, in order to directly supplement our main experimental findings, we ran multivariate regression for open-seat Senate races from 1960 to 2020, regressing our primary indicator of whether each party’s candidate was born in the state they ran in, along with the aforementioned list of controls, on the Democratic candidate’s two-party general election vote share statewide. We wish to acknowledge up front the potential role of selection effects in preventing completely definitive causal claims in the observational data—that is, we can only observe the general election outcomes for candidates who become their party’s nominee. In other words, the set of incumbent nominees with strong roots may not be comparable to the set of incumbent with weak local roots (see Ashworth, et al. 2024 for a similar issue in comparing male and female candidates). These effects present a potential for selection bias, although the signing of this bias is unclear.Footnote 17 Even so, these figures fulfill an important purpose of supplementing and supporting the core results of the survey experiment.

Nevertheless, the results found in Table 3 are highly consistent with our primary findings in the survey experiment. The survey experiment found evidence of about a 7-percentage-point increase in predicted vote share enjoyed by Democratic candidates with local credentials in open-seat races when facing a less-rooted Republican opponent. The observational model here predicts a 6.3 percentage point swing in vote share from an election in which only the Democratic candidate is locally born, to an election in which only the Republican is locally born.

Table 3 Effects on democratic vote share, open-seat senate candidates (1960–2020)
Full size table

The findings from our survey experiment also pointed to significant partisan asymmetry with regard to the advantages of local ties, with Democrats benefitting more—and more consistently—from these ties in eyes of their voters. The results in Table 3 suggest the same asymmetry: the size of the coefficient for Democrats’ electoral benefit from in-state birth (p < 0.01) is more than twice as large than the Republican counterpart (p < 0.1). These findings provide supplementary evidence of a substantial advantage for local candidates, particularly given the increased competitiveness associated with open-seat races rather than incumbent-challenger pairings.

We also observe advantages for incumbents, though they are notably less than their open-seat counterparts.Footnote 18 First, locally-rooted incumbents are significantly more likely to run unopposed in their general elections. Predicted probabilities generated from the results in Appendix Table A3 tell us that although the least-rooted Senate incumbents virtually never run without a challenger (probability of 0.1%), the deepest-rooted among them actually have a small chance (about 5%) of avoiding one entirely.Footnote 19 In the vast majority of incumbent Senate races, however, the opposing party does field a candidate. In these races, we find that local incumbent roots continue to be valuable. As Table 4 illustrates, the LRI produces a statistically significant coefficient of 0.63, which adds to a total effect (from the least to the most rooted incumbents) of around 2.5 points of vote share (with 95% confidence of a total effect between half a point and 5 points).

Table 4 Effects on two-party vote share, senate incumbents (1960–2020)
Full size table

The relatively small size of these effects for incumbents may not be wholly surprising. Incumbents have a much wider variety of criteria on which voters can judge them, including their records of achievement in Congress, their “home styles” as a serving member, and the association of their name with other national figures. Longer-serving members may come to be more associated with “Washington” and thus judged on a more nationalized scale. As a result, it is notable that local ties—which for incumbents were dug in very long ago—have any measurable impact for incumbents.

Finally, although our observational findings are mainly complementary to the main experimental survey findings, they can help close the gap on one shortcoming of experiment: that it represents just a snapshot in time of local candidate advantages. Our observational data, however, can summarize local advantages for both open-seat and incumbent candidates over a sixty-year period. We find no evidence that the beneficial effects of local roots on candidate electoral success have meaningfully diminished since 1960. The descriptive summaries in Fig. 5 demonstrate no obvious pattern of growing or diminishing electoral advantages for local candidates across the past 60 years of congressional elections. Senate candidates’ local ties today appear to be every bit as impactful as they were half a century ago. This is particularly notable given the concurrent rise of nationalized politics, party loyalty as an influence on voting behavior, and decline in influence of state and local parties and institutions during the same period.

Fig. 5
figure 5

Average two-party vote share advantage for locally-born candidates (compared to those born out-of-state) by decade, among open-seat candidates (black bars) and incumbent candidates (grey bars)

Full size image

Discussion

Former House Speaker Tip O’Neill famously argued that “all politics is local.” Although our political culture has nationalized and polarized since O’Neill’s heyday, our evidence suggests that at least some of American politics and representation are still very much local. We leverage a novel survey experiment, which suggests with high levels of certainty that candidates for the Senate are handsomely rewarded by the voters for sharing local place backgrounds and connections with their voters; and that those candidates without such connections are electorally punished for it. The effect of this local discrepancy is consistent across all analyses: between 2 and 7 additional points of vote share on average for open-seat candidates. We have shown that local biography and connections are unequivocally desirable as a trait for any candidate looking for additional support among the voters in their jurisdiction. Far from being a thing of the past, local roots are as meaningful and powerful a candidate attribute (at least from the voters’ perspective) as they have ever been, even in our current moment of deep polarization, partisan divisions and nationalized issue politics.

Our experiment has also provided novel evidence that the effects of local candidate roots are nuanced in how they intersect with party affiliation and attachment, on both the voter and candidate sides of the equation. First, we observe that local roots effects were larger among independents and opposite-party voters, suggesting that “place” and localness are unique identities that cross-cut partisanship. Relatedly, we have uncovered clear differences between the two major parties: although localness plays a major role for both parties, Democrats appear to benefit more from local connections, while Republicans face greater punishment for carpetbagging. Future work should more deeply investigate these asymmetric effects as they pertain to the two parties’ reputations among the voters—for example, whether this is due to growing urban/rural polarization between the parties, or perhaps to deeper-set cultural differences that are borne out in representational values.

In addition to our findings on local candidate preferences, we also uncovered novel descriptive trends not just about the local attributes of Senate candidates, but also about what voters assume about them. For example, carpetbagging is not an uncommon critique in elections at all levels of government and has been investigated qualitatively in compelling fashion in the Senate in particular (Galdieri, 2019). Yet when asked post-treatment about the survey’s believability, several respondents expressed doubt that any viable candidate for U.S. Senate would not have been born in the state they were running to represent. Imagine these respondents’ surprise were they to discover that a third of incumbents, and more than half of open-seat Senate candidates, were born and/or raised in another state, as our supplementary observational data reveals.

This leads to a broader and more important conclusion about the American electoral system: voters still often demand a local connection from their representatives (and punish them when they lack it) even as politics has clearly and substantially supplied nationalized brands. The American political tradition of hailing favorite sons and daughters—and of flogging carpetbaggers—is not new, but it does persist. Despite the chamber’s nationalizing trends, the Constitution still requires Senate both candidates to be “inhabitants” of their home states. Most Americans are almost certainly in favor of provisions like these. Although voters are often vocally supportive of “outsiders”, this almost always refers to those outside of the traditional political class, rather than literal outsiders from a geographic area.

A slew of important questions remain about local candidate connections and representation that deserve further attention. First, future work could explore differences in which local candidate credentials matter most. In our observational analysis of incumbents, we included all four local roots indicators (birthplace, high school, college, and postgraduate school) in a single additive index. Breaking these four indicators out into four separate models, each assessing the impact of the next in turn, tells us that all four indicators have broadly positive valence for Senate incumbents, but that birthplace and high school attendance appear to be the most individually meaningful.Footnote 20 This may be a function of what candidates discuss on the campaign trail about their biographies. It may also be that birthplace and high school—because they are largely out of a candidates’ control—speak to a greater sense of personal authenticity, and truly get at what it means to “grow up”; or be “born and raised” in a particular area. These markers, rather than later-in-life roots dug in after the fact, may in fact be more symbolically meaningful exhibitions of place attachment and local candidate connections.

Future research on this subject should also make further efforts to clearly isolate the mechanisms behind the power of local candidate roots. Previous scholarship has offered a number of potential frameworks—for example, Hunt’s practical/symbolic advantages distinction, Munis and others’ place identity concept, and our own theory of cross-cutting place representation. However, causal analysis that helps to determine which of these theoretical frameworks are most impactful on voters’ attitudes towards local and non-local candidates is important for expanding the study of local roots beyond elections and towards other elements of political representation, including whether the presence of local ties as an attribute is a major factor in legislative behavior.

Finally, this search for a clear mechanism would benefit from further exploration of potential heterogeneity in local candidate effects. It may be that these attachments play more of a role only in certain types of elections, particularly at the federal level, when the locus of attention is so often on presidential performance and national-level legislative activity in Washington, DC. Most notably, the novel asymmetry between the parties in the manner and extent to which their candidates benefit from local ties deserves further investigation, and in particular whether the two parties’ “brands” and reputations make one or the other more of a clear beneficiary of place-based voter support.